1. Could you explain the significance of your article to the non-specialist? 

Density functional theory (DFT) is the most popular computational method in chemistry, but problems are encountered in applications that involve both bond rearrangements and van der Waals interactions, e.g catalytic transformations of organic substrates in binding pockets of enzymes or on solid surfaces. Widely used functionals such as PBE do not properly account for long-range dispersion effects, but have been applied in many studies of hydrocarbon conversions in zeolite catalysts. We show that MP2 corrections to reaction energy profiles obtained with PBE (or other functionals) are substantial and not constant along the reaction path. By design our hybrid MP2:DFT approach is suited for studying reactions between a small or medium sized substrate molecule and a very large chemical system. 

2. What has motivated you to conduct this work? 

Hydrocarbon transformations in zeolites are an important class of industrial catalytic reactions and understanding their mechanisms will help to develop better catalysts and improved catalytic processes. There are also questions of fundamental chemical interest. One of them is whether or not the tert-butyl cation could exist as an intermediate in zeolites. Attempts to identify this species spectroscopically have not been successful so far. Computational studies are very instrumental in determining the structure and stability of intermediates in zeolites. However, techniques that are now routinely applied to typical zeolite simulation cells with several hundred atoms, either rely on force fields or on density functional theory. The former have problems with bond breaking and making processes, while the latter have problems with dispersion interactions. The method of choice, second order Moller-Plesset perturbation theory (MP2), cannot be applied to systems of this size. Therefore, we developed the MP2:DFT hybrid approach. 

3. Where do you see this work developing in the future? 

There are four directions of future developments: (i) We will see more applications of this method to other model cases of hydrocarbon reactions in zeolites, but also to different systems such as interactions of water molecules with fullerenes. (ii) The results obtained for this and other systems (estimates of the "periodic" MP2 limit) will serve as data base for parametrizing the damped dispersion corrections to DFT (Grimme and others). (iii) Application of the latter will be computationally less demanding and we will see many mechanistic studies of hydrocarbon reactions in zeolites by the "DFT+damped dispersion" method. Reliable reaction energy profiles will become available for hydrocarbon transformations in zeolites. (iv) Alternative approaches based on density functional theory will further develop and eventually become applicable to sub-systems which are connected by chemical bonds and, at the same time, interact also via dispersion forces. 

4. Are there any particular challenges facing future research in this area? 

Presently, reliable calculations of hydrocarbon-zeolite interactions (as of other large substrate-catalyst complexes) by hybrid MP2:DFT (or DFT+damped dispersion, for which the parameters are still lacking) are so expensive that one can localize minima or transition structures on the potential energy surface. Atomistic Monte Carlo or Molecular Dynamics simulations provide a more complete sampling of the configuration space, but rely on force fields and frequently ignore bond making/bond breaking processes at the reaction site. For example, the important effect of shape selectivity in zeolite catalysis has been studied by atomistic Monte Carlo simulations using empirical force fields (B. Smit et al., Phys Rev Lett 2005, 95, 164505; J. Catal. 2006, 237, 278) but bond formation at the active site has been neglected. Reactions at the active site were included in a hybrid DFT:force field approach, but configuration sampling was limited to the harmonic oscillator approximation (Clark & Sauer, J. Am. Chem. Soc. 2004, 126, 936). Convergence of these two developments is a challenge in future research. The question is: Can we calculate the potential energy surfaces fast enough (while keeping reliability) to allow for a more complete sampling of the potential energy surface by advanced Monte Carlo or Molecular Dynamics techniques?